Non-permissible Mobile Detection to Enhance Security in Bangladeshi Museum: A Multiprocess YOLO-Based Approach

Non-permissible mobile device use places Bangladeshi museums at risk for theft and resource devastation. Implementing deep learning, especially the YOLO can assist in detecting and preventing unauthorized mobile device use, improving Bangladeshi museum security. A deep learning model called Non-Permissible Mobile Detection in Museums (NMDM) can be used to identify faces, phones, and camera regions in real time. The proposed (NMDM) model analyzes n frames from real-time monitoring video, where museums deploy CCTV for security. This NMDM model, which is based on YOLOv8, predicts object detection (such as the identification of mobile, camera, and face) while also using ensemble approaches to decide whether to detect non-permissible mobile devices. To train that model here we apply our NMDM dataset by capturing photos of the phone in different angles and different conditions, after augmentation, the overall dataset size is 10494. In terms of testing results the NMDM model displays remarkable precision (97.671%) and accuracy (95.180%). With multi-processing, it also significantly speeds up both CPU and GPU systems.